Stress Test Analysis
Eight economic crisis scenarios were presented to digital personas, measuring four dimensions of customer behaviour
435 digital personas were each presented with 8 detailed crisis scenarios and asked four questions: Would you sell your investments? Would you withdraw your deposits? Would you switch banks? And how much do you still trust your bank? (1–5 scale). That's 13,920 questions in total (435 × 8 × 4), each with full qualitative reasoning, enabling us to understand not just what people would do, but why.
The scenarios range from plausible near-term events (regional real estate collapse, corporate debt defaults) to extreme "black swan" events (sovereign default, AI-driven trading losses, cyberattacks). This spread reveals which crises are truly dangerous for customer retention — and which barely register.
Four behavioural metrics reveal which scenarios are truly dangerous. Each stacked bar shows the proportion of yes (would act), maybe, and no responses.
The Liquidity Squeeze (Scenario D) is by far the most dangerous scenario: 67% would withdraw deposits and 51% would switch banks entirely. Nearly every customer (97.7%) would either withdraw or consider it. This scenario alone could trigger a self-reinforcing bank run.
Crises that don't directly threaten customer deposits — real estate collapse, energy debt, and digital asset crashes — show remarkable stability. Over 80% of customers would stay put. Customers distinguish between systemic threats to their money vs. broader market turbulence.
Mean trust score (1–5 scale) for each scenario. Higher = more trust retained. The gap between best and worst reveals the trust fragility spectrum.
There is a 1.45-point gap between the highest-trust scenario (Digital Asset Crash: 3.47) and the lowest (AI Trading Loss: 2.02). This is not a gradual decline — trust clusters into three tiers: resilient (3.2–3.5), strained (2.6–2.8), and broken (2.0–2.4). The broken-trust scenarios share a common trait: customers feel the bank directly caused or failed to prevent harm to their money.
Percentage who would sell their investments under each scenario
AI Trading Loss triggers the highest investment sell-off (40.3%) — even more than the Liquidity Squeeze. Customers see AI-caused losses as a direct failure of the bank's investment management, making them rush to liquidate. By contrast, a Cyberattack (2.8%) barely moves the needle on investments: customers worry about account access, not asset values.
All four metrics side-by-side for each scenario. Colour intensity indicates severity: red = high risk, green = low risk.
| Scenario | Withdraw Deposits % yes |
Switch Banks % yes |
Sell Investments % yes |
Trust Score mean /5 |
|---|---|---|---|---|
| D. Liquidity Squeeze | 66.9% | 50.8% | 33.3% | 2.38 |
| H. AI Trading Loss | 57.4% | 27.7% | 40.3% | 2.02 |
| F. Sovereign Default | 43.9% | 7.8% | 30.0% | 2.65 |
| C. Cyberattack | 24.5% | 26.3% | 2.8% | 2.73 |
| G. Pandemic Collapse | 20.3% | 20.5% | 11.4% | 3.23 |
| B. Energy Debt Default | 7.8% | 1.1% | 18.1% | 3.17 |
| E. Digital Asset Crash | 9.3% | 1.6% | 6.0% | 3.47 |
| A. Real Estate Collapse | 5.2% | 1.8% | 8.9% | 3.42 |
Scenarios cluster into three tiers. Tier 1 — Critical (D, H): multiple metrics in the red zone, trust below 2.5, majority of customers ready to act. Tier 2 — Elevated (C, F, G): one or two metrics alarming, trust strained but not broken. Tier 3 — Stable (A, B, E): customers distinguish between market turbulence and threats to their own deposits.
How each generation responds differently to the same crises. The gap between Baby Boomers and Millennials reveals fundamentally different risk psychology.
| Scenario | Baby Boomers | Gen X | Millennials | Gen Z |
|---|---|---|---|---|
| E. Digital Asset Crash | 3.95 | 3.40 | 3.33 | 3.32 |
| A. Real Estate Collapse | 3.81 | 3.40 | 3.24 | 3.39 |
| G. Pandemic Collapse | 3.77 | 3.17 | 3.05 | 3.00 |
| B. Energy Debt Default | 3.57 | 3.06 | 3.07 | 3.10 |
| C. Cyberattack | 3.11 | 2.68 | 2.63 | 2.57 |
| F. Sovereign Default | 3.04 | 2.55 | 2.59 | 2.45 |
| D. Liquidity Squeeze | 2.67 | 2.34 | 2.31 | 2.25 |
| H. AI Trading Loss | 2.27 | 1.93 | 1.99 | 2.03 |
| Scenario | Baby Boomers | Gen X | Millennials | Gen Z |
|---|---|---|---|---|
| D. Liquidity Squeeze | 54.2% | 68.1% | 72.4% | 68.8% |
| H. AI Trading Loss | 39.3% | 63.1% | 59.2% | 65.6% |
| F. Sovereign Default | 29.8% | 40.8% | 51.3% | 59.4% |
| C. Cyberattack | 9.4% | 26.4% | 28.7% | 32.3% |
| G. Pandemic Collapse | 3.5% | 21.2% | 26.6% | 25.0% |
| A. Real Estate Collapse | 0.0% | 5.0% | 6.3% | 15.6% |
| E. Digital Asset Crash | 1.2% | 10.6% | 13.3% | 6.2% |
| B. Energy Debt Default | 3.6% | 10.0% | 7.6% | 9.4% |
A closer look at the crises that could trigger genuine customer flight
A Liquidity Squeeze is the only scenario where nearly every customer would take action. With 97.7% either withdrawing or considering it, and over half ready to leave entirely, this scenario represents a genuine systemic threat. The self-reinforcing nature of bank runs means even "maybe" responses could rapidly convert to "yes" as customers see others withdrawing.
AI Trading Loss produces the lowest trust score (2.02) and highest investment sell-off (40.3%) of any scenario. Customers view AI-caused losses as a betrayal of fiduciary duty — the bank chose to use unproven technology with their money. Unlike a liquidity squeeze which feels external, AI loss feels self-inflicted by the bank, making the trust damage deeper and harder to repair.
Visualising all 8 scenarios across normalised risk dimensions. Larger area = greater overall customer impact.
Position reveals the nature of each threat. Scenarios in the upper-right combine high withdrawal rates with low trust — the most dangerous combination. Bubble size reflects bank-switching intent. Scenario H (AI Trading Loss) sits highest on the chart with the lowest trust score, while Scenario D (Liquidity Squeeze) pushes furthest right with the highest withdrawal rate and largest bubble. The safe scenarios cluster in the bottom-left: high trust, low withdrawal.
Stress-testing 10 major banks reveals a stark divide between traditional institutions and fintechs
Mean trust score (1–5). Green = high trust retained, red = trust collapsed.
| Bank customers |
A Real Est. |
B Energy |
C Cyber |
D Liquid. |
E Crypto |
F Sov. |
G Pandem. |
H AI Loss |
|---|---|---|---|---|---|---|---|---|
| Citi (31) | 3.71 | 3.58 | 3.00 | 2.45 | 3.81 | 2.94 | 3.65 | 2.06 |
| Schwab (23) | 3.68 | 3.45 | 3.22 | 2.38 | 3.82 | 2.65 | 3.57 | 1.77 |
| Chase (83) | 3.64 | 3.46 | 3.00 | 2.64 | 3.66 | 2.78 | 3.46 | 2.17 |
| Cap One (68) | 3.61 | 3.35 | 2.89 | 2.46 | 3.79 | 2.67 | 3.47 | 2.09 |
| Discover (32) | 3.60 | 3.28 | 2.81 | 2.33 | 3.55 | 2.58 | 3.43 | 1.81 |
| BofA (71) | 3.58 | 3.32 | 2.85 | 2.57 | 3.64 | 2.74 | 3.39 | 2.09 |
| Wells Fargo (55) | 3.44 | 3.13 | 2.78 | 2.42 | 3.52 | 2.85 | 3.23 | 2.08 |
| Chime (81) | 3.00 | 2.86 | 2.42 | 2.06 | 3.10 | 2.35 | 2.85 | 1.79 |
| PayPal (74) | 2.97 | 2.79 | 2.32 | 1.96 | 2.99 | 2.31 | 2.75 | 1.72 |
| Cash App (61) | 2.90 | 2.64 | 2.16 | 1.86 | 2.81 | 2.12 | 2.54 | 1.71 |
A clear dividing line separates traditional banks from fintechs. Cash App, PayPal, and Chime customers start with lower trust (2.9–3.0 even in mild scenarios) and it collapses further under stress: Cash App drops to 1.71 in the AI Trading Loss scenario vs. Chase at 2.17. These customers have weaker institutional loyalty and are far more likely to flee.
Percentage who would definitely withdraw. Red = high flight risk.
| Bank | A Real Est. |
B Energy |
C Cyber |
D Liquid. |
E Crypto |
F Sov. |
G Pandem. |
H AI Loss |
|---|---|---|---|---|---|---|---|---|
| Cash App | 16.4% | 16.4% | 46.7% | 83.6% | 24.6% | 65.6% | 42.6% | 75.4% |
| PayPal | 13.5% | 14.9% | 43.8% | 83.8% | 21.6% | 56.8% | 41.9% | 73.0% |
| Chime | 9.9% | 12.3% | 37.5% | 82.5% | 17.3% | 55.6% | 37.0% | 71.6% |
| Discover | 0.0% | 6.2% | 21.9% | 81.2% | 0.0% | 51.6% | 18.8% | 62.5% |
| Schwab | 0.0% | 9.1% | 8.7% | 78.3% | 0.0% | 54.5% | 17.4% | 73.9% |
| Cap One | 2.9% | 5.9% | 23.9% | 69.7% | 2.9% | 49.2% | 19.1% | 60.3% |
| Wells Fargo | 1.8% | 3.6% | 18.2% | 67.3% | 10.9% | 37.0% | 25.5% | 56.4% |
| Chase | 7.2% | 6.0% | 19.5% | 66.7% | 9.6% | 38.3% | 15.7% | 45.1% |
| Citi | 0.0% | 6.7% | 12.9% | 65.5% | 0.0% | 41.9% | 6.5% | 51.6% |
| BofA | 4.2% | 7.1% | 23.9% | 64.8% | 8.5% | 50.0% | 16.9% | 54.9% |
In the AI Trading Loss scenario, Cash App customers withdraw at 75.4% vs. Chase at 45.1% — a 30-point gap. Even in milder scenarios like Cyberattack, the fintech-to-traditional gap persists: 47% of Cash App users would withdraw vs. 20% of Chase users. Traditional bank customers show consistently more patience and willingness to wait things out.
Across the dangerous scenarios, 20–35% of respondents say "maybe" rather than committing to action. Analysing their reasoning reveals five factors that would push them over the edge:
1. Duration of disruption. The most cited trigger. "Maybe" respondents consistently say they'd wait days, not weeks. If access issues or bad news persist beyond a short window, they convert to action. The crisis itself isn't the tipping point — the bank's speed of resolution is.
2. Loss of fund access. ATM limits, frozen transfers, or app outages are the single fastest converter. Even customers who trust their bank will withdraw if they physically cannot reach their money. Access anxiety overrides institutional loyalty.
3. Poor communication. Silence from the bank during a crisis pushes "maybe" to "yes." Respondents who mention proactive bank communication are more likely to stay. Those who describe overwhelmed support lines or no updates describe tipping toward withdrawal.
4. Seeing others act. Several respondents describe a social trigger: if friends, family, or news coverage shows other customers withdrawing, they'd follow. This is the self-reinforcing dynamic that turns individual hesitation into a bank run.
5. Switching friction is the main brake. Direct deposits, autopay, and the sheer hassle of moving banks are repeatedly cited as the reason people stay in "maybe" rather than acting. This friction is a genuine retention asset — but it only delays, not prevents, flight if the underlying issue isn't resolved.
Which banks' customers are most likely to leave entirely or liquidate investments under stress?
| Bank | A Real Est. |
B Energy |
C Cyber |
D Liquid. |
E Crypto |
F Sov. |
G Pandem. |
H AI Loss |
|---|---|---|---|---|---|---|---|---|
| Schwab | 13.0% | 0.0% | 21.7% | 73.9% | 0.0% | 31.8% | 21.7% | 31.8% |
| Discover | 3.1% | 0.0% | 31.2% | 68.8% | 6.2% | 12.5% | 28.1% | 37.5% |
| Chime | 1.2% | 1.2% | 33.3% | 65.4% | 1.2% | 6.2% | 29.6% | 36.2% |
| Cash App | 3.3% | 0.0% | 44.1% | 65.0% | 1.6% | 9.8% | 26.7% | 43.3% |
| PayPal | 1.4% | 0.0% | 41.7% | 60.3% | 4.1% | 9.6% | 32.9% | 38.4% |
| BofA | 2.9% | 0.0% | 30.9% | 51.4% | 0.0% | 8.7% | 11.3% | 29.9% |
| Wells Fargo | 3.6% | 0.0% | 29.6% | 50.0% | 1.9% | 9.3% | 23.6% | 25.9% |
| Chase | 3.6% | 0.0% | 22.2% | 48.8% | 1.2% | 8.4% | 22.9% | 24.7% |
| Cap One | 4.4% | 0.0% | 16.7% | 48.5% | 3.0% | 13.2% | 14.7% | 26.9% |
| Citi | 3.2% | 0.0% | 13.3% | 48.4% | 0.0% | 9.7% | 12.9% | 33.3% |
Charles Schwab customers show the highest switching intent in the Liquidity Squeeze: 73.9% would leave, compared to 48–49% for Chase, Capital One, and Citi. Schwab's investment-focused customer base is both more financially engaged and more willing to move. Their 56.5% investment sell-off rate in Scenario D is also the highest of any bank.
| Bank | A Real Est. |
B Energy |
C Cyber |
D Liquid. |
E Crypto |
F Sov. |
G Pandem. |
H AI Loss |
|---|---|---|---|---|---|---|---|---|
| Schwab | 21.7% | 30.4% | 8.7% | 56.5% | 9.1% | 52.2% | 13.0% | 45.5% |
| BofA | 11.3% | 25.7% | 5.7% | 53.5% | 4.3% | 35.2% | 9.9% | 36.2% |
| Discover | 12.5% | 15.6% | 0.0% | 37.5% | 3.2% | 40.6% | 6.2% | 53.1% |
| Cash App | 6.7% | 18.0% | 6.6% | 33.3% | 4.9% | 33.3% | 16.4% | 49.2% |
| Wells Fargo | 7.3% | 24.1% | 5.7% | 48.1% | 5.5% | 29.1% | 12.7% | 46.3% |
| Citi | 12.9% | 19.4% | 0.0% | 45.2% | 0.0% | 35.5% | 9.7% | 43.3% |
| Chime | 8.8% | 15.0% | 3.7% | 35.0% | 6.2% | 28.7% | 17.5% | 43.8% |
| PayPal | 6.8% | 13.5% | 5.5% | 33.8% | 8.1% | 28.8% | 14.9% | 41.9% |
| Cap One | 8.8% | 17.6% | 1.5% | 36.8% | 2.9% | 33.3% | 13.2% | 38.5% |
| Chase | 15.7% | 19.3% | 1.2% | 34.1% | 8.6% | 27.7% | 12.2% | 35.8% |
What these results mean for bank resilience planning
The single most important finding: customers rarely flee over market turbulence — they flee when they doubt access to their own money. Of eight scenarios tested, only those that directly threatened deposit access or trust in the bank's competence (Liquidity Squeeze, AI Trading Loss, Sovereign Default) produced majority withdrawal intent. Banks should build their crisis resilience around this principle: protect perceived deposit safety above all else, communicate proactively about access guarantees, and be especially transparent about any AI-driven decision-making that touches customer assets.